Exploring Shoreline Change in North West Sierra Leone

This post aims to analyze the rates of shoreline change in the Scarcies region of North West Sierra Leone along the Atlantic Ocean coastline. The analysis will utilize data from the Digital Earth Africa program Digital Earth Africa program to gain valuable insights into the evolving patterns of coastal movement over time.

Visualize the data

The following image is a screenshot showing the visualization of the data through the interactive map displayed below.

The map below has the following features:

  • The area of interest is highlighted in yellow.

  • Data points are represented by different colors, indicating shoreline behavior over time. Please refer to Table 1 for the color key and cluster counts.

  • To navigate the map and select specific layers, utilize the interactive buttons provided.

  • On the left-hand section, you’ll find a “layers button” that allows easy navigation between different data sets.

  • Removing the “World” layer improves the level of detail in the satellite images provided by the tmap() function.

  • The extent of mangroves is displayed in green pixels. (Make sure to deactivate the “World” layer)

Important Insights

  1. Challenging Coastline and Model: The coastline and the model used to measure shoreline distances are complicated due to factors like image noise, currents, deltas, and islands. I suggest divide the region into smaller parts and clean the data for each area individually. This will help analyze the shoreline changes in more detail.

  2. Outliers: Values greater than 200 meters are more likely to indicate modeling issues rather than real-world coastal change.

  3. Noisy data points: The data set contains 3,920 points categorized as noisy. Additionally, in the year 2001, there are 1,398 fuzzy data points, which have been removed entirely.

  4. Median values per year: Using the median values, a linear model suggests an erosion rate of 0.304 meters per year, indicating a potential loss or erosion of 30.4 centimeters annually.

  5. Mean values per year: Running a linear model on the mean values indicates that for every one unit increase in year, the mean shoreline distance is projected to grow by 1.4 meters.

Questions:

  1. Should the focus be on data points that show erosion related to sea-level rise, rather than a broader shoreline analysis?

  2. Are we interested on the correlation between shoreline movement and the presence of mangroves?

  3. Should I use a function like Winsorize() to control outliers (> 200 ) data points by setting maximum and minimum points for shoreline erosion?

Data Appendices

Table 1

The table 1 establishes color clusters on the interactive map based on the rate_time, which represents annual rates of change (in meters per year). These values are calculated by Digital Africa using linear regression models that analyze shoreline distances over time (excluding outliers). Negative values indicate retreat, while positive values indicate growth.

Table 1. Data Points Color Clusters
colors lower_bound upper_bound count
darkred -Inf -5.00 110
red -5.00 -1.00 514
pink -1.00 -0.25 1033
white -0.25 0.00 712
white 0.00 0.25 712
lightblue 0.25 1.00 407
blue 1.00 5.00 288
darkblue 5.00 Inf 173

Histograms

The histograms are an overview of the data:

  • Each data point in the histogram represents the shoreline’s distance in a specific year compared to the shoreline position in 2021.

  • Negative values indicate shoreline growth, indicating that the shoreline moved further inland compared to 2021.

  • Positive values indicate shoreline erosion or movement away from the reference point over time.

  • The data points in the earlier years are more scattered and show higher volatility. As we approach 2021, the number of data points decreases.

  • Some years exhibit a greater number of data points above and below the 200-meter mark. These extreme values are flagged as potentially noisy based on the metadata provided.

OLS Model

## 
## Call:
## lm(formula = median ~ year, data = .)
## 
## Coefficients:
## (Intercept)         year  
##    617.0338      -0.3042
## [1] "The result of -0.3042 using the median values year by year indicates that the land is likely eroding at a rate of approximately 0.304 meters per unit of time (e.g., per year). The lm() function in R fits a linear model (OLS)."
## 
## Call:
## lm(formula = mean ~ year, data = .)
## 
## Coefficients:
## (Intercept)         year  
##   -2821.586        1.398
## [1] "The result of 1.398 using the mean values year by year suggests that the land is likely growing at a rate of approximately 1.398 meters per unit of time (e.g., per year). This positive value indicates a direction of coastal growth, with the land expanding over time."

Linear Model Graph

Annual Shoreline Change Rates

Webpage vs. OLS Analysis by Year

This code is comparing the rate_time values obtained from a webpage with those obtained from an OLS analysis for each year, to identify any differences between the two.

Table 2

Table 2. Model comparison
Data Point rate_time linear_slope
e9wd7gce4v -0.49 -0.5390554
e9wd7gcfch -0.34 -0.3204402
e9wd7gcfxk -0.40 -0.3646631
e9wd7gexz4 -1.21 -1.2477487
e9wd7gezt6 -1.03 -1.0639324
e9wd7gf3ge -0.89 -0.8567742 
## [1] 0.2033859
## [1] 0.1919374

This table and results provides a comparison between the rate_time variable obtained from the digital analysis and the manually computed linear model lm().

The table 2, shows the first 6 rows of results exhibiting the rate_time value from Digital Africa and the linear_slope value calculated with the lm() function. To evaluate the model’s overall fit, I computed the mean rate_time value, which equaled 0.2033 meters. Furthermore, the linear slope analysis generated a value of 0.1919, which closely aligns with the mean rate_time.